George Batchkala

Which normalization transform whould I use to use pre-trained models?

4

Hi, I want to use your pre-trained ResNet models, but I could not find which transforms you used. To be more precise, when using ImageNet pre-trained models on new inputs,...

Make the selection of critical instance consistent with https://github.com/binli123/dsmil-wsi/blob/1e8f11128843e65438c9609267d6ad2d6d190692/train_tcga.py#L47 **Test that the results are the same:** 

Training: give bags (one slide = one bag) in a different order every epoch * Shuffling was present but unused (line 35) Testing: remove shuffling since it does not make...

Thank you very much for making the blog post and sharing the code! Can you please explain the code in "3. Data distribution VS Stats distribution"? I do not understand...

- TIA Toolbox version: 1.5.1 - Python version: 3.11.8 - Operating System: CentOS Linux 7 (Core) ### Description In the example notebooks, the files are sometimes downloaded into the `examples`...

Add support for `ignore_is_tiled_tiff` in `WSIReader.open()`

3

A possible solution for the issue https://github.com/TissueImageAnalytics/tiatoolbox/issues/804 **Modification**: Added a flag `ignore_is_tiled_tiff` to the `open()` method of `WSIReader` from `tiatoolbox/wsicore/wsireader.py` **Usecase**: if you know that openslide worked well on your...

`.tif` file from Ventana scanner is not parsed properly with WSIReader; openslide works well on its own

2

- TIA Toolbox version: 1.5.1 - Python version: 3.11.8 - Operating System: CentOS Linux 7 (Core) ### Description I have slides scanned with a **Ventana DP200 scanner and saved as...

Weights are saved twice after running the walkthrough notebook: `models--MahmoodLab--UNI` and `models--MahmoodLab--uni`

2

Dear authors, I ran the [walkthrough notebook](https://github.com/mahmoodlab/UNI/blob/main/notebooks/uni_walkthrough.ipynb), and the model was downloaded twice (see screenshot). 1. I get the symbolic link `assets/ckpts/vit_large_patch16_224.dinov2.uni_mass100k/pytorch_model.bin` pointing to a file in `.cache/huggingface/hub/models--MahmoodLab--UNI/`. 2. The...

Dear authors, `faiss` is unnecessary for using the UNI model unless someone wants to use `uni/downstream/eval_patch_features/protonet.py` and the corresponding part of the walkthrough notebook. I added the link to the...